Fully Parallelized Multi-prover Protocols for NEXP-Time

Abstract

A major open problem in the theory of multi-prover interactive proofs is to characterize the languages which can be accepted by fully parallelized multi-prover protocols with an exponentially low probability of cheating. In this paper we solve this problem by proving that any language which can be accepted by a sequential multi-prover protocol can also be accepted by a single-round multi-party protocol, and thus the multi-prover round hierarchy collapses to its first level: MIP(poly)=MIP(1)=NEXP-time.